云—雾—边缘协同的数字孪生制造系统仿真过程动态扰动响应方法  

Cloud-fog-edge collaborative digital twin manufacturing system simulation process and dynamic disturbance response method

作　　者：孟麒 胡天亮[1,2,3] 马嵩华 MENG Qi;HU Tianliang;MA Songhua(School of Mechanical Engineering,Shandong University,Jinan 250061 China;Key Laboratory of High-Efficiency and Clean Mechanical Manufacture(Shangdong University),Ministry of Education,Jinan 250061,China;National Demonstration Center for Experimental Mechanical Engineering Education,Jinan 250061,China)

机构地区：[1]山东大学机械工程学院,山东济南250061 [2]高效洁净机械制造教育部重点实验室,山东济南250061 [3]机械工程国家级实验教学示范中心,山东济南250061

出　　处：《计算机集成制造系统》2023年第6期1996-2005,共10页Computer Integrated Manufacturing Systems

基　　金：国家重点研发计划资助项目(2020YFB1708403);国家自然科学基金资助项目(51875323);山东省重点研发计划资助项目(2019JZZY010123-1)。

摘　　要：鉴于现有仿真技术对制造过程动态响应能力不足,数字孪生技术为仿真系统对制造过程的动态映射提供了解决方案,但仍存在设备运行过程中动态扰动识别困难且扰动识别模型的训练过程受制于计算和数据资源,导致耗时较长,以及用于识别动态扰动和构建约束条件的数据为离线或半离线状态,不能满足实时响应要求的问题。针对以上问题,提出一种基于云—雾—边缘协同的数字孪生制造仿真过程动态扰动响应方法。在云端使用公共数据集训练普适化模型并使用边缘端数据训练个性化模型来提升扰动识别精度,雾端通过部署云端训练的个性化模型保证扰动识别速度,同时将识别的扰动更新到数字孪生仿真模型中。边缘端负责采集实时信号并上传至雾端。经实验验证,扰动约束更新机制能够准确更新扰动,从而对运行过程中的扰动做出快速响应。Modeling and simulation has become one of the core technologies in manufacturing industry.However,the existing simulation technology is insufficient for the dynamic response of the manufacturing process.Digital twin provides a solution for dynamic mapping of the simulation system to manufacturing process.At present,there are still existing the following problems:①it is difficult to recognize dynamic disturbances in the process of equipment operation,and the training process of disturbance identification model is limited by computing and data resources,which is time-consuming;②data utilized for identification of dynamic disturbance and update of constraints are offline or semi-offline,which can not meet the requirements of real-time response.To solve the above problems,a method for responding to dynamic disturbances in digital twin manufacturing simulation process based on cloud-fog-edge collaboration was proposed.In the cloud,a universal model was trained by public data sets,and a personalized model further was trained by edge data to improve the accuracy of disturbance recognition.In the fog,the personalized model was deployed to ensure the speed of disturbance recognition,while the recognized disturbance was updated to the digital twin simulation model.In the edge end,real-time signals were collected and uploaded to the fog end.The experimental results showed that the disturbance constraint updating mechanism could accurately update the disturbance and make a quick response to the disturbance during operation.

关 键 词：数字孪生 扰动识别 卷积神经网络 云—雾—边缘协同 

分 类 号：TH133.3[机械工程—机械制造及自动化] TP206.3[自动化与计算机技术—检测技术与自动化装置]